🐛 Describe the bug As the title, when using onnx to export a quantized convolution layer, the outcome will have plus or minus one difference in some positions with the quantized convolution layer. ...

ONNX will act as the model export format in PyTorch 1.0 and will allow for the integration of accelerated runtimes or hardware-specific libraries.

According to Facebook, PyTorch 1.0 takes the modular, production-oriented capabilities from Caffe2 and ONNX and combines them with PyTorch's existing flexible, research-focused design to provide a ...

So trt2 is about 20% faster than trt1. I suspect the presence of explicit quantization nodes in trt1 might be causing the slowdown. Is this expected behavior? Does doing PTQ in the ONNX space generate ...

Alternatively, PyTorch 1.0 integrates the capabilities from Caffe2 and ONNX and combines it with PyTorch's ability to provide a seamless path from research prototyping to production deployment.